考慮CpG島及生物路徑之基因體DNA甲基化貝氏階層模型

Abstract

近年來表基因體學(epigenetics)逐漸受到重視，在這些不涉及任何基因型(genotype)改變，卻會影響生物表現型(phenotype)的機制中，DNA甲基化(DNA methylation)是表基因體學中最為蓬勃發展的一類。DNA甲基化在維持細胞的正常功能及調節基因與環境的相互作用上扮演重要的角色，透過甲基化�去甲基化調控DNA轉錄，進而影響細胞分化及基因表現。因此，DNA甲基化很適合做為疾病的生物標記(biomarker)，為疾病的診斷、預後評估、治療藥物開發提供新的研究方向。過去的研究指出DNA甲基化的狀態與其是否位於CpG島(CpG islands)有關，在正常的細胞中，CpG島通常處於非甲基化狀態，而CpG島以外的區域通常處於甲基化狀態；當細胞發生癌變，CpG島之甲基化程度異常地升高 (hypermethylation)，而CpG島以外的甲基化程度異常地降低 (hypomethylation)。除此之外，曾有研究指出DNA甲基化與生物路徑(biological pathway)有關，在生物路徑中相鄰的基因在疾病狀態下傾向同時發生異常甲基化�去甲基化。然而過去探討DNA甲基化的研究中，沒有研究者同時考慮CpG島及生物路徑的影響。因此我們提出一基因體DNA甲基化貝氏階層模型，同時納入CpG島及生物路徑訊息。為瞭解此方法的表現，我們進行一卵巢癌的實際資料分析。實際資料分析的結果顯示，我們提出的統計模型能夠偵測出與疾病相關的生物路徑及其中扮演重要角色的基因。總結來說，本文提出的統計模型能夠偵測出與疾病相關的生物路徑，以用來探討致癌機轉；同時在考慮生物機制的情況下，偵測出最有可能實際應用在臨床的生物標記。

DNA methylation (DNAm) occurs at cytosines in CpG dinucleotides and alters gene expression, and hence it may lead to development of diseases. DNAm is an intermediate factor between genome sequence, environmental factors and gene expression; therefore, it could serve as a good choice of biomarker. Both CpG islands (CGI) and biological pathway might have effects on DNA methylation. However, among the methods for studying differential DNAm, to the best of our knowledge, no one has considered incorporating these two factors into analysis simultaneously. We proposed a Bayesian hierarchical model for differential DNAm profiling that takes these two factors into consideration: (1) The CpG location in CGI or not. (2) Pathway information of the gene that the CpG located. We considered six different functions to describe pathway information. The functions concern the presence of genes in pathways, the number of neighboring genes in pathways, and the CpG located in CGI or not. To evaluate the performance of the proposed model, we analyzed the DNAm data of United Kingdom Ovarian Cancer Population Study (UKOPS). The results showed that our proposed model can provide a novel insight into the identification of the disease-related pathways and disease-related genes contained in the pathways. In conclusion, our approach could serve as a tool to explore the molecular mechanism of the development and progression of disease; moreover, based on the pathway information, it may help identify the biomarkers with potential use in clinical applications.